On Oct 27, 8:32 pm, Carl Banks <[EMAIL PROTECTED]> wrote: > I was wondering if anyone had any advice on this. > > This is not to study graph theory; I'm using the graph to represent a > problem domain. The graphs could be arbitrarily large, and could > easily have millions of nodes, and most nodes have a substantial > amount of data associated with them. Obviously I don't want a whole > such graph in memory at once, so libraries the just deal with in- > memory graphs are out. > > I know I could implement this with a relational DB, and I'd be fine > with a library that was built on top of one. But I'm hoping for > specialzed behavior useful for graphs.
If you're looking for FOSS, the Boost graph library [1] or its parallel extension [2] is probably your best bet; it also comes with Python bindings but they are not maintained any more. For commercial solutions, Star-P [3] seems an interesting platform, with bindings to Matlab and Python. Freebase [4] is apparently built on a special graph database but unfortunately only the stored data are available, not the DB source code. George [1] http://www.boost.org/doc/libs/1_36_0/libs/graph/doc/index.html [2] http://www.osl.iu.edu/research/pbgl/ [3] http://www.interactivesupercomputing.com/success/sparsematrices.php [4] http://www.freebase.com/help/faq#q7 -- http://mail.python.org/mailman/listinfo/python-list
